Terry Blanton <[email protected]> wrote: To pass the Turing test the 'bot will have to perceive the world as a human: >
I doubt a computer will be able to pass a "no rules" Turing test anytime in the next ~100 years. That is, when the test allows the person to type in any question or comment about any subject. I think that computers will fumble questions about relate to being a human being in a human body. I mean things about sex, love, death, jealousy, food, dieting and so on. However, if you restrict the Turing test to a subject area, such as you do with a tech support customer response, then a computer might pass the test within 20 years. A human customer support technician might be allowed to banter or make small talk for a moment -- "how's the weather in Atlanta?" -- but after that they have to stick to the question at hand. They are not allowed to talk about sex. A computer could simulate that restriction even now. Siri does a pretty good job at it. Another interesting prospect will be the reverse Turing test. Will people learn to imitate computers so well they can fool other people into thinking they are computers? That should not be so difficult. People opposed to cold fusion such as Mary Yugo are already well along in that art. I sometimes suspect she is a version of the computer therapist Eliza. Supposedly, some of the early movies of things like railroad locomotives heading straight for the camera frightened audiences. People nowadays are used to movies and they never mistake them for reality, even momentarily. I expect people will gradually become used to artificial intelligence, and they will become better at recognizing intelligent computers 50 years after they first emerge. In some cases, we might be fooled by computers because we have low expectations of them. If you could show a Siri cellphone to a computer expert in 1955, he might say: "That is a person talking like a computer. No real computer could come up with such remarkable answers." - Jed
